From: "William A. (Andy) Adamson" Subject: Re: Am I still the one to fix the NFS_FILE_SYNC problem Date: Mon, 24 May 2010 12:47:09 -0400 Message-ID: References: <4BF975A3.7090509@panasas.com> <4BFA9D9D.2030908@panasas.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Cc: Benny Halevy , NFS list To: Boaz Harrosh Return-path: Received: from mail-gw0-f46.google.com ([74.125.83.46]:53796 "EHLO mail-gw0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757687Ab0EXQrL convert rfc822-to-8bit (ORCPT ); Mon, 24 May 2010 12:47:11 -0400 Received: by gwaa20 with SMTP id a20so1578847gwa.19 for ; Mon, 24 May 2010 09:47:10 -0700 (PDT) In-Reply-To: <4BFA9D9D.2030908@panasas.com> Sender: linux-nfs-owner@vger.kernel.org List-ID: On Mon, May 24, 2010 at 11:39 AM, Boaz Harrosh w= rote: > On 05/24/2010 04:14 PM, William A. (Andy) Adamson wrote: >> Hi Boaz >> >> The last patch I sent in [PATCH 1/1] SQUASHME pnfs-submit: replace >> layoutcommit_ctx with rpc_cred fixes the NFS_FILE_SYNC problem for >> files. It replaces taking a reference on the open context with takin= g >> a reference on the rpc_cred. >> > > Hi Andy. > > I hope you are online could you chat with me a bit about this, > on boazharrosh-Re5JQEeQqe9fmgfxC/sS/w@public.gmane.org > I have seen your patch and i like it a lot. I have not yet test with > it, so I own you. > > I'm talking about the other problem I had where the write-out returns > NFS_FILE_SYNC and commit is not called and pnfs_client_ops->nfs_commi= t_complete() > is not eventually called, then inode never gets written/updated and w= e get all > these straining .nfs000000000000XXX files and IO never finishes. You = said > you had them too and I had a work around for objects. (This problem i= s new > in 2.6.34) > > Do you have this problem or it's only me? Yes, I had that problem - but for the file layout it was the pnfs_need_layoutcommit at the end of pfns4_write_done that was preventing _nfs4_close from being called due to the get on the nfs_open_context. Replacing the nfs_open_context with the rpc_cred solved our problem. Also, remember that the file layout does not need a layout commit - it's just a hint. -->Andy > > Boaz > >> -->Andy >> >> >> On Sun, May 23, 2010 at 2:36 PM, Boaz Harrosh = wrote: >>> >>> Andy, Benny are you waiting on me to fix the NFS_FILE_SYNC problem? >>> >>> I've been on a long holiday, but I can attempt a fix tomorrow if st= ill >>> needed? >>> >>> I was hoping someone else need that more urgently then me, since I = already >>> have a work around, for my testing ;-) >>> >>> Boaz >>> >>> -- >>> To unsubscribe from this list: send the line "unsubscribe linux-nfs= " in >>> the body of a message to majordomo@vger.kernel.org >>> More majordomo info at =A0http://vger.kernel.org/majordomo-info.htm= l >>> > >